If a physical event p has a cause c occurring at a time t, then p has a physical cause p’ such that 1) p occurs at t and 2) p is a supervenience base of c.
This principle could be given a justification similar to the one we gave Closure-Overdetermination: it could be argued that if the principle is not true, absurdly pervasive overdetermination will follow (unless mental events happen to be identical to physical events—a possibility we will consider later). For lack of space, we leave working out the exact justification for Closure-Supervenience, as well as how Kim’s second argument could be completed using it, as an exercise for the reader.
17 Again, this dichotomy may not be exhaustive, but the argument goes through no matter what kind of property we consider in place of M.
18 Again, if F is an n-place relation, x is an ordered n-tuple of objects.
19 Namely, the property that Kim calls the event's “constitutive” property (as explained in section 7 below).
20 Perhaps more accurately, the conjunction of token physicalism with Irreducibility, which is stronger than the denial of type physicalism. The falsity of type physicalism would only require the existence of one property that is not a physical property.
21 As Fancesco Orilia also notes in his contribution to this volume (section 6). He also points out that Kim (1996, 60) explicitly recognizes this implication of his account of events.
22 There are others that appeal to the idea of “realisation” or “constitution”, and that are compatible with a fine-grained conception of events. Cf. Boyd (1980, 82-87; 101-103), Cummins (1983, 22-23); Papineau (1993, ch. 1).
23 Unsurprisingly, since on his fine-grained conception of events, physicalism must minimally require that mental events, albeit distinct from physical events, somehow “depend on” and be “determined by”, physical events. And “realisationist” versions of nonreductive physicalism obviously imply supervenience, since the realisation relation implies the supervenience relation (as Kim recognizes in his 1998, 23-24).
24 As one would expect under a coarse-grained conception of events.
25 On p. 44 Kim indicates that he would not identify Brutus’s stabbing Caesar with Brutus’s killing Caesar, and on p. 43 he says that “no stabbings are killings and no killings are assassinations”. These are similar enough to our examples.
26 On p. 44 Kim concedes that Brutus’s stabbing Caesar and Brutus’s stabbing Caesar with a knife are the same event and he accordingly denies that the relation x stabs y with a knife is a constitutive property of the event denoted by “Brutus’s stabbing Caesar with a knife”, though it clearly is a property of the pair Brutus, Caesar.
27 Kim intends his “existence condition” schema to be understood with the restriction that only predicates that ascribe constitutive properties may be substituted for “P” (Kim 1976, 34-37). Our own attitude about what events exist is liberal: we do not think our existence schema “The event in which x has P at t exists iff x has P at t” (section 2) is in need of restricting. Events in which properties of every kind are instantiated are required for the semantic analysis of natural language. Only some events can enter into the causal relation—only the metaphysically contingent ones, we suppose, viz. events e such that it is metaphysically possible both for e to have occurred and for e to not have occurred. For discussions of causation, our existence schema could be restricted so that only predicates that ascribe contingent properties (properties such that possession of them is a metaphysically contingent matter) may be substituted for “P”, and the semanticist’s events could be called something other than “events” (“schmevents”, for example), but this is a question of word choice, not substance. Such a restriction is intelligible inasmuch as the notion of metaphysical possibility is, and, it seems to us, it would correctly disqualify as events those schmevents that cannot enter into the causal relation.
Note that the question whether to restrict the schema
(E1) The event in which x has P at t exists iff x has P at t,
which we used in section 2, is different from that of whether to restrict the schema
(E2) [x, P, t] exists iff x has P at t,
which is what Kim is concerned with in his 1976. The latter is a partial definition of a new piece of notation (“[”, “]”), call this the canonical event description notation. Ruling out a particular predicate “R” as an allowable substituend for “P” in (E2) would not have the consequence that there is no such thing as the event in which x has R at t; it would only have the consequence that that event, if it exists, is not denoted by “[x, R, t]”—it might be denoted by some other canonical event description. Presumably Kim would not impose the same restriction on the ordinary language (E1) as on the quasi-formal (E2), since in 1976 he wants to allow events to instantiate properties which are not constitutive of them (see our note 24). The restriction Kim proposes for (E2) differs from the one just contemplated for (E1) in that the former is not made using terms of which we have a prior understanding but by using the technical term “constitutive property”, which Kim does not explicitly define (see pp. 36-37).
28 Kim notes that “a revision of the standard property-exemplification account of events (essay 3) [Kim 1976]” may be called for, “especially if mental properties, in spite of their multiple realisability, are accepted as legitimate event-generating properties. For on the standard account two property instances count as distinct events if the properties instantiated are distinct ... . Considerations advanced in Kim (1992) concerning disjunctive properties may be reason enough for excluding mental properties as constitutive properties of events” (Kim 1993, 364-65, note 5, our italics).
29 The closest Kim comes to providing an argument is in the following: “Suppose that a certain event, in virtue of its mental property, causes a physical event. The causal closure of the physical domain says that this event must also have a physical cause. We may assume that this physical cause, in virtue of its physical property, causes the physical event. The following question arises: What is the relationship between these two causes...?” (1989, 280). Kim goes on to suggest that, barring overdetermination and given closure, the physical cause excludes the mental one unless the mental and the physical properties of the two causes are identified. The implication here is that the mere “token identity” of two causes as coarse-grained events won’t give us a solution to the problem; what needs to be identified is “that in virtue of which” the cause causes what it does.
30 This is essentially the principle attributed to Kim by Marras, 2000, p 145, as principle (Q*). Note that is addition to a quausal exclusion principle, the reconstructed argument would also require analogous revisions of Closure and/or of SC I and SC II, as well as of any premise to the effect that a mental (physical) event causes another event. (E.g., Closure would become Quausal Closure: “If a physical event e has a cause that occurs at time t, then it has a physical cause c that occurs at t and that causes e in virtue of some physical property of which c is an instance”.)
31 Which is itself far from obvious. See Marras 1998 for discussion.
32 See Macdonald and Macdonald 2006, 566, Bennett 2003, and Papineau 1993, ch. 1, sections 6 and 7.
33 This was, essentially, Kim’s (1993) complaint against Davidson’s anomalous monism. While granting that Davidson was “arguably right” in denying that anomalous monism entails the causal inertness of mental properties, he nonetheless insists that anomalous monism “fails to provide mental properties with a causal role” (p. 20). As argued in Marras (1997), this was merely a failure of omission: Davidson’s monism was not intended to provide a theory of the causal efficacy of mental properties.
34 See, e.g., Fodor 1989, McLaughlin 1989, Jackson and Pettit 1990, Hardcastle 1998.
35 Kripke’s well-known modal arguments against the identity theory, for example, can equally be directed against the type and the token identity theory. So can any of the familiar arguments based on the conceivability/possibility of zombies and/or disembodied minds. Such arguments must of course be distinguished from arguments against nonreductive (versus reductive) physicalism, either of the token identity variety or of the realisation variety. These latter arguments need to show not that token identity is false, but that the denial of type identity is false. Typically these aim to show that the multiple realisability thesis is false, or that it does not stand in the way of a reductive account of mental properties. See, e.g., Kim 1992, Richardson 1979, Bickle 1998.
36 For a recent, forceful critique of such arguments see Papineau 2002.
37 Consider, for example Burge’s (1979) argument against token physicalism. The argument is something like this: Tokens of those mental states that are broadly individuated cannot be identical to tokens of the physical states that realize them because the realizing properties are plausibly intrinsic ones. Suppose, for example, that Joe’s belief that p at time t is realized by Joe’s brain state S (“Joe’s S” for short). If the content of Joe’s belief is broadly individuated, then there is a possible world w in which Joe is intrinsically exactly like he is in the actual world, so he has S at t in w, but in which Joe’s does not believe that p at t because of some difference in his environment. Now it would seem that on the basis of this we can argue:
It is possible Joe’s belief that p occurs but Joe’s S does not occur (premise).
It is possible that Joe’s belief that p ≠ Joe’s S (from (1) by Leibniz’s law and propositional modal logic).
Joe’s belief that p ≠ Joe’s S (from (2) because objects that are possibly distinct are actually distinct).
The argument can clearly be reformulated so as to apply not just to Joe’s (token) belief that p and to Joe’s S, but also to the types (properties) believing that p and being in S. After all, “It is possible that Joe’s belief that p occurs but…” is necessarily equivalent to “It is possible that believing that p is exemplified by Joe but…”
That having been said, we note that the argument is not particularly convincing when applied (as intended) against token physicalism. That is, while it is true that if the argument is successful in refuting token physicalism, it is also successful in refuting type physicalism, we do not think that the argument is successful against token physicalism. The trouble is that, as a matter of modal logic, in order for the move from (1) and (2) to (3) to be valid, or, equivalently, in order for ‘a ≠ b a ≠ b’ to be a logical truth, ‘a’ and ‘b’ must be rigid designators. However, the token events at issue in the argument are denoted by definite descriptions, which are not rigid—at least not rigid de jure. Definite descriptions may, however, be rigid de facto if they happen to pick out their designata by their essential properties (e.g. ‘the even prime’ is rigid de facto). Burge of course recognizes this—he claims that tokens of beliefs have their contents essentially (Burge 1979, 75), but this is far from obvious. In fact, the very externalist considerations Burge brings to bear on token physicalism seem to militate against this conclusion. It seems to us entirely natural to describe the moral of Burge (1979) by saying things like: “There are counterfactual circumstances in which Joe’s belief that p at t would have had a different content”, “If the stuff in the oceans had been XYZ, then (the present token of) my belief that water is wet would have had a different content than it in fact does”. (Note that the argument would be valid if the descriptions in (1) were read as having wide scope relative to the modal operator. However, if (1) were read this way, it would receive no support from Burge’s externalist considerations: the latter only show that the mental properties of a person can come apart from the physical properties that, in the actual world, realize them. They do not show that the events in which these properties are instantiated in the actual world are distinct in some other world.)
If, on the other hand, the argument is made using definite descriptions that designate properties instead of events, there is no problem about rigidity. Any property—at least any property we have a predicate for—can be designated by a rigid definite description that makes use of the predicate we would normally use to ascribe the property. For example “the property of being red”, or “the property red” are rigid: in no world does “the property of being red” designate any property other than redness. Similarly, “the property of believing that p” in no world designates anything other than the property of believing that p.
38 That Kim thinks that the problem of mental causation poses special difficulties is evident in ch. 3 of his 1998 book, especially in view of his claim that the supervenience argument “does not generalize” to other domains beyond the mental.
39 For a similar stance and a thorough discussion of the issues, see Ladyman and Ross 2007, ch. 5.
40 However, there are some promising developments: recent attempts to explicate causation in terms of the outcomes of counterfactual interventions (Woodward 1997, Pearl 2000) might provide us with the means to do just this.